Convolutional Neural Network for OCT Interferogram Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing OCT systems face challenges in accurately detecting subtle changes in retinal fluid pooling due to complex and high-resolution requirements, and the difficulty in comparing interferograms from different scan patterns, which limits the effectiveness of fluid detection and monitoring.
Innovation Solution
A system and method using a trained convolutional neural network (CNN) to process interferogram images from OCT data, where the CNN is trained on data from one scan pattern and applied to data from another pattern, with data interpolation, extrapolation, or resampling to enhance recognition of retinal tissue layers and fluid pooling, allowing for more precise measurements and comparison across different scan patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution OCT systems are used to detect retinal fluid pools, then detection accuracy is improved, but device complexity increases and availability is limited
Solution Approach 1:
A trained CNN model serves as an intermediary between standard-resolution OCT systems and fluid pool detection. The model processes interferogram images from conventional OCT systems to identify fluid pools, eliminating the need for complex high-resolution hardware while maintaining detection accuracy.
Solution Approach 2:
The patent replaces complex mechanical/optical high-resolution OCT systems with a computational approach using trained neural networks. The CNN model substitutes physical hardware complexity with software-based image processing to achieve equivalent or superior fluid pool detection.
2Productivity
If interferograms from different scan patterns are compared directly, then comparison speed is maintained, but measurement accuracy deteriorates due to pattern differences
Solution Approach 1:
The patent transforms interferogram images by adjusting parameters such as resolution, contrast, and geometric properties to normalize different scan patterns. This allows direct comparison of interferograms from various scan patterns while maintaining measurement accuracy through consistent parameter representation.
3Reliability
If subtle changes in gray scale values are analyzed manually, then diagnostic thoroughness is improved, but detection capability worsens due to human limitations
Solution Approach 1:
The trained CNN model performs self-service by automatically analyzing interferogram images to detect fluid pools and subtle gray scale changes. The system eliminates the need for manual analysis while enhancing detection capability through the model's ability to recognize patterns beyond human visual perception.
Solution Approach 2:
The patent replaces manual visual inspection with automated neural network analysis. The CNN model substitutes human diagnostic capability with computational image processing that can detect subtle gray scale variations and fluid pool characteristics more reliably than manual methods.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The solution enables more accurate and efficient detection of retinal fluid pooling and thickness measurements, improving the monitoring of eye health and disease progression, and facilitating the comparison of scans from different OCT systems with varying scan patterns.
Implementation Method 1
Some interferometers function by splitting light from a single source into two beams that travel in different optical paths, and are then combined again to produce the interference patterns.
Data Source
AI summary
A neural network is trained to segment interferogram images. A first plurality of interferograms are obtained, where each interferograms corresponds to data acquired by an OCT system using a first scan pattern, annotating each of the plurality of interferograms to indicate a tissue structure of a retina, training a neural network using the plurality of interferograms and the annotations, inputting a second plurality of interferograms corresponding to data acquired by an OCT system using a second scan pattern and obtaining an output of the trained neural network indicating the tissue structure of the retina that was scanned using the second scan pattern. The system and methods may instead receive a plurality of A-scans and output a segmented image corresponding to a plurality of locations along an OCT scan pattern.